Sentiment Classification with Support Vector Machines and Multiple Kernel Functions

Abstract

Support vector machine (SVM) is a learning technique that performs well on sentiment classification. The performance of SVM depends on the used kernel function. Hence, if the suitable kernel is chosen, the efficiency of classification should be improved. There are many approaches to define a new kernel function. Non-negative linear combination of multiple kernels is an alternative, and the performance of sentiment classification can be enhanced when the suitable kernels are combined. In this paper, we analyze and compare various non-negative linear combination kernels. These kernels are applied on product reviews to determine whether a review is positive or negative. The results show that the performance of the combination kernels that outperforms the single kernels.